透析计算:面向OLGP的InfoNetCube高效物化

发布时间：2018-05-14 15:44

本文选题：信息网络 + 部分物化　；参考：《软件学报》2017年03期

【摘要】：信息网络数据立方(InfoNetCube)的计算是进行信息网络在线分析处理的基础.然而,不同于传统的数据立方,信息网络数据立方由多个子方体格组成,每个方体格中任意方体(cuboid)的任意单元格都包含一个主题图(或称图度量),因而空间开销较传统数据立方大2个数量级以上.如何快速、高效地进行信息网络数据立方的部分物化,是极具挑战的研究课题.提出了基于透析计算思想的信息网络立方物化策略,通过主题图度量在信息维和拓扑维上反单调性运用,提出了基于透析计算的空间剪枝算法,快速透析掉不可能命中的子图度量、方体单元、方体乃至方体格.实验结果表明,所提出的基于透析计算的部分物化策略可以对信息网络方体进行有效剪枝,算法较基于基本方体的部分物化策略运行时间平均降低75%.
[Abstract]:The calculation of information network data cube (InfoNet Cube) is the basis of on-line analysis and processing of information network. However, unlike traditional data cubes, information network data cubes consist of multiple child cubes, Any cell in each cube contains a topic graph (or graph metric), so the space cost is more than two orders of magnitude larger than the traditional data cube. How to implement partial materialization of information network data cube quickly and efficiently is a challenging research topic. The information network cubic materialization strategy based on the idea of dialysis computing is proposed. The spatial pruning algorithm based on dialysis computation is proposed by using the topic map metric in the information and topological dimension. Quickly dialysis out the impossible sub-graph metric, square unit, square and even square physique. The experimental results show that the proposed partial physicochemical strategy based on dialysis computation can effectively prune the information network square, and the average running time of the partial physicochemical strategy based on the basic cube is reduced by 75%.
【作者单位】：四川大学计算机学院;工业和信息化部电子第五研究所;
【基金】：国家自然科学基金(61103043) 国家科技支撑计划(2012BAG04B02)~~
【分类号】：TP311.13

【相似文献】